Linear model of coregionalization : http://www.jmlr.org/papers/volume12/alvarez11a/alvarez11a.pdf. More...

#include <multioutput_kernels.hpp>

Inheritance diagram for gplib::multioutput_kernels::lmc_kernel:

Collaboration diagram for gplib::multioutput_kernels::lmc_kernel:

Classes
struct	implementation

Public Member Functions
	lmc_kernel ()
	Constructor. More...

	lmc_kernel (const std::vector< std::shared_ptr< kernel_class >> &kernels, const std::vector< arma::mat > &params)
	Constructor, requires the inner kernels to be used and the parameter matrices. More...

	lmc_kernel (const size_t lf_number, size_t n_outputs)
	Constructor, requires the number of latent functions and the number of outputs to be used, creates default kernels and parameters. More...

	~lmc_kernel ()
	Destructor. More...

arma::mat	eval (const std::vector< arma::mat > &X, const std::vector< arma::mat > &Y, bool diag=false) const
	Evaluates the kernel function over the provided sets of matrices. More...

arma::mat	derivate (size_t param_id, const std::vector< arma::mat > &X, const std::vector< arma::mat > &Y, bool diag=false) const
	Returns the value of the derivative wrt a certain parameter with a a particular pair of input matrices. More...

size_t	n_params () const
	Returns the total number of parameters needed by the kernel (parameter matrices, plus the parameters of each inner kernel). More...

void	set_params_k (const std::vector< arma::mat > &params)
	Sets the parameters of the multioutput kernel only, doesn't affect the parameters of the inner kernels. More...

void	set_params (const std::vector< double > &params, size_t n_outputs=0)
	Sets all the parameters of the multioutput kernel including those of the inner kernels using a std. More...

void	set_param (size_t q, size_t a, size_t b, const double param)
	Sets the parameter of the matrix B in multioutput kernel. More...

void	set_param (size_t q, size_t param_id, const double param)
	Sets the parameter of the matrix B in multioutput kernel. More...

void	set_kernels (const std::vector< std::shared_ptr< kernel_class >> &kernels)
	Sets the inner kernels. More...

std::vector< arma::mat >	get_params_k () const
	Returns a vector of matrices containing the parameters of the multioutput kernel, but not those of the inner kernels (in other words only the parameter matrices). More...

std::vector< double >	get_params () const
	Returns a std vector containing all of the parameters of the multioutput kernel, including those of each inner kernel. More...

std::vector< std::shared_ptr< kernel_class > >	get_kernels () const
	Returns a shared pointer to a vector containing the inner kernels currently set. More...

double	get_param (size_t q, size_t a, size_t b) const
	Returns a double with a single parameter of the matrix B of the multioutput kernel. More...

double	get_param (size_t q, size_t param_id) const
	Returns a double with a single parameter in the inner kernel of the multioutput kernel. More...

void	set_lower_bounds (const double &lower_bounds)
	Sets the lower bounds to be used by the kernel during training process including those of the inner kernels. More...

void	set_upper_bounds (const double &upper_bounds)
	Sets the upper bounds to be used by the kernel during training process including those of the inner kernels. More...

void	set_lower_bounds (const std::vector< double > &lower_bounds)
	Sets the lower bounds to be used by the kernel during training process including thos of the inner kernels. More...

void	set_upper_bounds (const std::vector< double > &upper_bounds)
	Sets the upper bounds to be used by the kernel during training process including thos of the inner kernels. More...

std::vector< double >	get_lower_bounds () const
	Returns a vector with the lower bounds of all the parameters, including those of the inner kernels. More...

std::vector< double >	get_upper_bounds () const
	Returns a vector with the upper bounds of all the parameters, including those of the inner kernels. More...

Public Member Functions inherited from gplib::multioutput_kernel_class
	multioutput_kernel_class ()
	Multioutput Kernel Class definition. More...

	multioutput_kernel_class (const std::vector< std::shared_ptr< kernel_class >> &kernels, const std::vector< arma::mat > &params)
	Constructor, requires the inner kernels to be used and the parameter matrices. More...

virtual	~multioutput_kernel_class ()=default
	Destructor. More...

Private Attributes
implementation *	pimpl

Detailed Description

Linear model of coregionalization : http://www.jmlr.org/papers/volume12/alvarez11a/alvarez11a.pdf.

Constructor & Destructor Documentation

gplib::multioutput_kernels::lmc_kernel::lmc_kernel ( )

Constructor.

gplib::multioutput_kernels::lmc_kernel::lmc_kernel	(	const std::vector< std::shared_ptr< kernel_class >> &	kernels,
		const std::vector< arma::mat > &	params
	)

Constructor, requires the inner kernels to be used and the parameter matrices.

Parameters

kernels	: Shared pointer containing a vector with the inner kernels (kernel_class).
params	: Vector containing the parameter matrices.

gplib::multioutput_kernels::lmc_kernel::lmc_kernel	(	const size_t	lf_number,
		size_t	n_outputs
	)

Constructor, requires the number of latent functions and the number of outputs to be used, creates default kernels and parameters.

Parameters

lf_number	: size_t, Number of latent functions.
n_outputs	: size_t, Number of outputs.

gplib::multioutput_kernels::lmc_kernel::~lmc_kernel ( )

Destructor.

Member Function Documentation

mat gplib::multioutput_kernels::lmc_kernel::derivate	(	size_t	param_id,
		const std::vector< arma::mat > &	X,
		const std::vector< arma::mat > &	Y,
		bool	diag = `false`
	)		const

virtual

Returns the value of the derivative wrt a certain parameter with a a particular pair of input matrices.

Parameters

param_id	: Identifier of the parameter we are derivating with respect to.
X	: First vector of matrices for derivative evaluation.
Y	: Second vector of matrices for derivative evaluation.
diag	: Flag, if it is true the kernel should only be evaluated for the derivative entries pertaining to the diagonal of the answer matrix, this is due to performance reasons while using FITC.

Implements gplib::multioutput_kernel_class.

mat gplib::multioutput_kernels::lmc_kernel::eval	(	const std::vector< arma::mat > &	X,
		const std::vector< arma::mat > &	Y,
		bool	diag = `false`
	)		const

virtual

Evaluates the kernel function over the provided sets of matrices.

Parameters

X	: First vector of matrices for kernel evaluation.
Y	: Second vector of matrices for kernel evaluation.
diag	: Flag, if it is true the kernel should only be evaluated for the entries pertaining to the diagonal of the answer matrix, this is due to performance reasons while using FITC.

Implements gplib::multioutput_kernel_class.

vector< shared_ptr< kernel_class > > gplib::multioutput_kernels::lmc_kernel::get_kernels ( ) const

virtual

Returns a shared pointer to a vector containing the inner kernels currently set.

Implements gplib::multioutput_kernel_class.

vector< double > gplib::multioutput_kernels::lmc_kernel::get_lower_bounds ( ) const

virtual

Returns a vector with the lower bounds of all the parameters, including those of the inner kernels.

Implements gplib::multioutput_kernel_class.

double gplib::multioutput_kernels::lmc_kernel::get_param	(	size_t	q,
		size_t	a,
		size_t	b
	)		const

Returns a double with a single parameter of the matrix B of the multioutput kernel.

Parameters

q	identifier of the latent function.
a	identifier of the param row.
b	identifier of the param col.

double gplib::multioutput_kernels::lmc_kernel::get_param	(	size_t	q,
		size_t	param_id
	)		const

Returns a double with a single parameter in the inner kernel of the multioutput kernel.

Parameters

q	identifier of the inner kernel.
param_id	the identifier of the parameter.

vector< double > gplib::multioutput_kernels::lmc_kernel::get_params ( ) const

virtual

Returns a std vector containing all of the parameters of the multioutput kernel, including those of each inner kernel.

Implements gplib::multioutput_kernel_class.

vector< mat > gplib::multioutput_kernels::lmc_kernel::get_params_k ( ) const

virtual

Returns a vector of matrices containing the parameters of the multioutput kernel, but not those of the inner kernels (in other words only the parameter matrices).

Implements gplib::multioutput_kernel_class.

vector< double > gplib::multioutput_kernels::lmc_kernel::get_upper_bounds ( ) const

virtual

Returns a vector with the upper bounds of all the parameters, including those of the inner kernels.

Implements gplib::multioutput_kernel_class.

size_t gplib::multioutput_kernels::lmc_kernel::n_params ( ) const

virtual

Returns the total number of parameters needed by the kernel (parameter matrices, plus the parameters of each inner kernel).

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_kernels ( const std::vector< std::shared_ptr< kernel_class >> & kernels )

virtual

Sets the inner kernels.

Parameters

kernels : Shared pointer containing a vector of kernel_class kernels.

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_lower_bounds ( const double & lower_bounds )

virtual

Sets the lower bounds to be used by the kernel during training process including those of the inner kernels.

Parameters

lower_bounds : double, the lower bounds of all the parameters will be set to this value.

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_lower_bounds ( const std::vector< double > & lower_bounds )

virtual

Sets the lower bounds to be used by the kernel during training process including thos of the inner kernels.

Parameters

lower_bounds : Vector containing the lower bounds to be used.

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_param	(	size_t	q,
		size_t	a,
		size_t	b,
		const double	param
	)

Sets the parameter of the matrix B in multioutput kernel.

Parameters

q	identifier of the latent function.
a	identifier of the parameter row.
b	identifier of the parameter col.
param	the value of the parameter to be set.

void gplib::multioutput_kernels::lmc_kernel::set_param	(	size_t	q,
		size_t	param_id,
		const double	param
	)

Sets the parameter of the matrix B in multioutput kernel.

Parameters

q	identifier of the inner kernel.
param_id	identifier of the parameter.
param	the value of the parameter to be set.

void gplib::multioutput_kernels::lmc_kernel::set_params	(	const std::vector< double > &	params,
		size_t	n_outputs = `0`
	)

virtual

Sets all the parameters of the multioutput kernel including those of the inner kernels using a std.

vector.

Parameters

params	: vector containing all the parameters needed by the multioutput kernel.
n_outputs	: If passed a number bigger than 0 the parameter matrices will be shaped according to the number received.

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_params_k ( const std::vector< arma::mat > & params )

virtual

Sets the parameters of the multioutput kernel only, doesn't affect the parameters of the inner kernels.

Parameters

params : Vector of matrices containing the parameters the size of the vector should be the same as the number of latent functions (inner kernels).

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_upper_bounds ( const double & upper_bounds )

virtual

Sets the upper bounds to be used by the kernel during training process including those of the inner kernels.

Parameters

upper_bounds : double, the upper bounds of all the parameters will be set to this value.

Implements gplib::multioutput_kernel_class.

void gplib::multioutput_kernels::lmc_kernel::set_upper_bounds ( const std::vector< double > & upper_bounds )

virtual

Sets the upper bounds to be used by the kernel during training process including thos of the inner kernels.

Parameters

upper_bounds : Vector containing the upper bounds to be used.

Implements gplib::multioutput_kernel_class.

Member Data Documentation

implementation* gplib::multioutput_kernels::lmc_kernel::pimpl

private

The documentation for this class was generated from the following files:

src/gplib/multioutput_kernels.hpp
src/gplib/multioutput_kernels.cc

Classes

Public Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation